The end goal of research in natural language processing is to provide a psychologically plausible computational account of human language processing. To this end, cross-linguistic investigation of sentence processing can provide critically significant results which elucidate the interaction between the grammars of particular languages and the mechanisms of human language processing system. In the proposed research, one critical aspect of sentence processing, namely, how speakers of Japanese and English deal with temporal ambiguities that are encountered during on-line processing of a sentence. In particular, we ask how speakers of Japanese deal with a high degree of indeterminacy early in the sentence. In order to investigate whether Japanese speakers are engaged in extensive reanalyses en route to a correct interpretation, five types of sentences are proposed to be examined: sentences with relative clauses; sentences with stative verbs; passive and causative sentences; sentences with verbs of giving and receiving; and sentences with lexical homonymy. These sentences all contain structures that might trigger reanalysis. Syntactic reanalysis of grammatical relations as well as reanalysis of thematic role assignments will be examined. The sentences will be tested by three psycholinguistic experiments: eye-movement monitoring; self-paced reading methods; and on- line grammaticality judgement task. Two questionnaires concerning peoples' subjective judgments of conscious reanalysis and difficulty of the sentence will also be conducted to supplement the experimental data. The results of the Japanese studies will then be contrasted to English data (available in the literature) to ask whether the ways in which Japanese speakers process temporal ambiguity are similar to that of English speakers. Using these cross-linguistic results, we will address the question of whether a single mechanism can account for the way speakers of Japanese and English deal with temporal ambiguity, or alternatively, whether their differences are better accounted for by invoking two different parsing mechanisms.